TY - JOUR
T1 - Enhanced Detection of Glass Insulator Defects Using Improved Generative Modeling and Faster RCNN
AU - Ning, Pin
AU - Jin, Jin
AU - Xu, Yuanping
AU - Kong, Chao
AU - Zhang, Chaolong
AU - Tang, Dan
AU - Huang, Jian
AU - Xu, Zhijie
AU - Li, Tukun
N1 - Publisher Copyright:
© 2024 The Authors. Published by Elsevier B.V.
PY - 2024
Y1 - 2024
N2 - The precise defect detections for glass insulators are of utmost importance to ensure their safety and functionality. Therefore, this study proposes an improved defect detection algorithm based on the optimized deep learning network. The collected glass insulator data was limited in size. Even after data augmentation, it remained insufficient for the training requirements of deep learning models. In this study, employing the improved Denoising Diffusion Probabilistic Models (DDPM) generative model to expand the glass insulator data. The glass insulator defect images generated by the improved noise-fitting network exhibit enhanced quality and high fidelity. Through manual selection and iterative experiments, a total of 1200 images were curated, constituting the glass insulator defect dataset for training and test of deep learning models. Faster RCNN was chosen as the defect detection model, and its VGG16 feature extraction network was replaced with ResNet50 to address the issue of gradient vanishing caused by excessive network stacking. Additionally, the Feature Pyramid Network (FPN) structure was introduced to enhance semantic extraction for different defect scales. The Kmeans++ algorithm was utilized to improve the proposal box generation parameters in RPN. Compared to the baseline Faster RCNN model's mAP of 72.7%, our improved version achieved a significant increase, reaching a mAP of 85.9%.
AB - The precise defect detections for glass insulators are of utmost importance to ensure their safety and functionality. Therefore, this study proposes an improved defect detection algorithm based on the optimized deep learning network. The collected glass insulator data was limited in size. Even after data augmentation, it remained insufficient for the training requirements of deep learning models. In this study, employing the improved Denoising Diffusion Probabilistic Models (DDPM) generative model to expand the glass insulator data. The glass insulator defect images generated by the improved noise-fitting network exhibit enhanced quality and high fidelity. Through manual selection and iterative experiments, a total of 1200 images were curated, constituting the glass insulator defect dataset for training and test of deep learning models. Faster RCNN was chosen as the defect detection model, and its VGG16 feature extraction network was replaced with ResNet50 to address the issue of gradient vanishing caused by excessive network stacking. Additionally, the Feature Pyramid Network (FPN) structure was introduced to enhance semantic extraction for different defect scales. The Kmeans++ algorithm was utilized to improve the proposal box generation parameters in RPN. Compared to the baseline Faster RCNN model's mAP of 72.7%, our improved version achieved a significant increase, reaching a mAP of 85.9%.
KW - Convolutional neural network
KW - DDPM
KW - Deep learning
KW - Defect detetction
KW - Glass insulator
UR - http://www.scopus.com/inward/record.url?scp=85209626254&partnerID=8YFLogxK
U2 - 10.1016/j.procir.2024.10.007
DO - 10.1016/j.procir.2024.10.007
M3 - Conference article
AN - SCOPUS:85209626254
SN - 2212-8271
VL - 129
SP - 31
EP - 36
JO - Procedia CIRP
JF - Procedia CIRP
T2 - 18th CIRP Conference on Computer Aided Tolerancing, CAT 2024
Y2 - 26 June 2024 through 28 June 2024
ER -